Professional Documents
Culture Documents
ALGORITHM
Submitted by
SANDEEP S KURUP
Fourth Semester
M.Tech, Microwave and Television Engineering
Submitted by
SANDEEP S KURUP
Third Semester
M.Tech, Signal Processing
0
DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING
COLLEGE OF ENGINEERING
TRIVANDRUM
CERTIFICATE
This is to certify that this Thesis preliminary report entitled “DYNAMIC CHASE
MUD FOR MC-CDMA ” is a bonafide record of the Thesis preliminary presented
by Sandeep S. Kurup, under our guidance towards partial fulfillment for the award
of Master of Technology Degree in Electronics and Communication Engineering
(Signal Processing), of the University of Kerala during the year 2015.
Sandeep S. Kurup
i
ABSTRACT
ii
The Chase algorithm replaces the search along the whole lookup table (or
code), where every codeword is considered, with a search among a new much smaller
lookup table where only the least reliable positions of a code are replaced with their al-
ternatives. This dramatically reduces the complexity of the decoder while still offering
a near-ML performance.For a code of base m and length n , the length of the lookup
table is reduced from mn codes for the ML decoder to mp for the Chase decoder, where
p n is the number of least-reliable symbols to be varied.
The Chase decoder has parameters which can be varied and will result in
changing the system’s complexity. One parameter is the number of least reliable sym-
bols p which can be considered to create the lookup table. This can range from one
to the total length of the code in use, n , (the optimal ML filter case).The maximum
length of lookup table in a Chase decoder is mp , but this can be reduced with slight
degradation in the system’s performance.
iii
Contents
1 INTRODUCTION 1
2 LITERATURE REVIEW 12
6 CONCLUSION 34
iii
List of Figures
iv
Chapter 1
INTRODUCTION
The MC CDMA meets very high data rate requirements and inherits the properties
of Code Division Multiple Access (CDMA) and Orthogonal Frequency Division Mul-
tiplexing (OFDM) [1].MC CDMA has the advantages of multicarrier modulation, on
one hand, and the flexibility offered by the spreading technique, on the other hand. It
has got two major demerits, v.i.z. Multiple Access Interference (MAI) and high Peak
to Average Power Ratio (PAPR). The MAI is produced by the interfering signals of the
k − 1 users. So as the number of users increases, the effect of MAI becomes dominant.
The equalisation, interference cancellation and multiuser detection are the three basic
methods used for the elimination of the MAI in MC CDMA systems. The multiuser
detection schemes utilizes the signals of the k − 1 users for the joint detection of the
k th user’s signal. The optimum multiuser detector is the maximum likelihood receiver
finding the euclidean distances of the received vector with each code words in the ml
lookup table. As the number of users increases, this look up table becomes such huge
that the practical realization becomes impossible owing to the complexity.
There are a number of various modulation varieties in CDMA systems, such as DS-
CDMA in which spreading sequences are directly multiplied with information bear-
ing signals, FH-CDMA in which the carrier frequency changes with the spreading se-
quences, TH-CDMA in which the information bearing signals are transmitted in short
bursts according to spreading sequences instead of continuously, and hybrids of these
techniques. Among these different techniques, DSCDMA is the most popular since it
has several attractive properties for wireless medium [4], such as easy generation of
spreading codes, easy generation of frequency synchronizer, and possibility of coher-
ent demodulation. However, the traditional use of the DS-CDMA technology does not
appear realistic for very high data-rate multimedia services at the speeds of the order of
several hundred megabits per second due to the severe multipath-induced interchip and
intersymbol interference (ISI) as well as multiple access interference (MAI). [5]
The "orthogonal" part of the OFDM name indicates that there is a precise math-
ematical relationship between the frequencies of the carriers in the system. It is possible
2
Figure 1.1: OFDM signal generation and transmission
to arrange the carriers in an OFDM Signal so that the sidebands of the individual carri-
ers overlap and the signals can still be received without adjacent carrier’s interference.
In order to do this the carriers must be mathematically orthogonal. If the integral of the
product of two signals is zero over a time period, then these signals are orthogonal to
each other. The orthogonality allows simultaneous transmission on a lot of sub-carriers
in a tight frequency space without interference from each other. So in the receiver side
easily we can extract the individual sub-carriers. But in traditional FDM systems over-
lapping of carriers are not possible, rather a guard band is provided between each carrier
to avoid inter-carrier interference.
OFDM solves the problem of ISI through use of a cyclic prefix due to high data
rates. It also provides other advantages like high spectral efficiency, Low implementa-
3
tion complexity etc. Some of the major applications of OFDM include digital audio
broadcasting digital video broadcasting, local area networks, WiMax etc. Despite of all
these advantages and applications the carrier frequency offset (CFO) and high peak to
average power ratio (PAPR) are major disadvantage of OFDM. This disadvantage needs
to be addressed properly to allow further widespread use of OFDM.
The advantages of OFDM include high spectral efficiency, insensitivity to ISI and
ICI , higher data rate, better performance with less complexity and Computational ef-
ficiency.The major drawbacks of OFDM are [7] high Peak to Average Power Ratio
(PAPR) and high sensitivity to carrier frequency offset.
4
Figure 1.3: MC-CDMA Transmitter Receiver
For those multimedia applications which require very high data rate (up to several
hundred megabits per second), DS-CDMA is not feasible because of the severe ICI and
ISI at very high data rate. MC-CDMA was developed to alleviate the problem. MC-
CDMA was the combination of CDMA and OFDM techniques with an aim to enhance
wireless system capabilities and accommodate the benefits of both CDMA and OFDM
systems [2] and may even outperform both. The principle of MC-CDMA is to spread
the codes in the frequency domain: a single signal is replicated into L parallel copies;
each of the copies is multiplied by a single chip of a spreading code of length L and
then modulated (usually using BPSK modulation) to a subcarrier.
Besides the advantages of CDMA and OFDM, MC-CDMA also has the following
5
impressive advantages. It has much simpler receiver design compared with DS-CDMA.
Theoretically, a DS-CDMA system with spreading sequences of length N could accom-
modate N users simultaneously, but it is always infeasible because of the tremendous
computational load. A MC-CDMA system can handle these N users easily using a
standard receiver structure. Furthermore, MC-CDMA provides guard interval for each
symbol, so a RAKE receiver used to combat multipath effects is no longer necessary.
MC CDMA has got higher spectral efficiency. OFDM often uses coding techniques
to enhance the performance; as a result, the numbers of subcarriers are greater than
those of the symbols transmitted simultaneously. In MC-CDMA systems, there is no
coding needed at the transmitter, so the number of subcarriers is much less than that of
OFDM. MC-CDMA provides inherent frequency diversity in that each individual sub-
carrier may encounter fading independently. Using diversity combining techniques may
enhance the system performance. However, MC-CDMA still has the following major
drawbacks; [7] high Peak to Average Power Ratio (PAPR) and Sensitivity to frequency
offset and phase noise. Frequency offset and phase noise will damage orthogonality
between subchannels therefore will cause the system performance lose.
Multiple access interference is the major factor that limits the capacity and
performance of DS-CDMA systems. Earlier detectors did not consider MAI and per-
formed single user detection. To mitigate MAI effects, those detectors focused on
improving spreading code waveform design, power control, forward error correction
(FEC) code design, sectored/adaptive antenna design, and so on [9]. This approach
is simple and easy to implement. However, it ignores the crosscorrelation between
spreading sequences and the (possibly) known structure of the data signals. Therefore,
the performance of this scheme is poor and vulnerable to the near-far problem. To
combat the near-far problem and MAI, more sophisticated detection strategies, such
as multiuser detection (MUD), need to be implemented at the receiver. Multiuser de-
tection is a strategy for mitigating MAI effects at the receiver side. For the last ten
years, various multiuser detection schemes have been developed to combat the effects
of MAI. Most early research work was focused on conventional multiuser detectors,
6
which assume knowledge of all active users, including both known users and unknown
interfering users, perform interference cancellation and achieve optimal or suboptimal
detection results. For example, with the knowledge of all active users’ spreading se-
quences, the classic decorrelating detector can completely eliminate the effects of MAI
at the expense of enhancing background noise. This multiuser detection scheme offers
superior performance compared to the traditional decorrelating detectors [10]; however
it is also complex. The complexity increases dramatically with the number of users
which makes it hard to implement in practice. Meanwhile, conventional multiuser de-
tectors [11] require too much information about users, such as the knowledge of all the
spreading sequences, received signal amplitudes, and so on, which makes them unre-
alistic in practice. Compared to decorrelating detection, blind multiuser detection has
been developed with only knowledge of the desired user(s) information so that such
designs may be much more practical. Though blind multiuser detection has no a pri-
ori knowledge of other users’ information, they normally use additional statistic signal
processing procedures to estimate other users’ spreading sequences or channel informa-
tion [9]. This normally takes more time or bandwidth; otherwise, detection performance
may be sacrificed.
7
of intense research on multiuser detection.
The goal of linear MUD is to attain as much of the capacity increase from
optimum MUD as possible, with a feasible and low-complexity implementation based
on linear filters. [13] A linear multiuser detector is simply a filter that is designed to
attenuate MAI according to a specific criterion. There are two important types of linear
multiuser detectors, known as decorrelating detector, and the minimum mean square
error (MMSE) detectors.
i Decorrelating Detector
A dual mode multiuser detector with a switch between matched filter detector
and decorrelator is used where switching action based on channel conditions decide the
path between two detectors so that it is neither using matched filter nor decorrelator all
the time, keeping average performance.
ii MMSE Detector
8
The drawback of standard MMSE receiver is that it cannot track the change
of the channel parameters such as amplitude and phase of the fading process. [13] To
overcome this drawback, there have been various efforts on MMSE receiver in fading
channels.This receiver improves performance by compensating channel phase varia-
tions. But the performance is still degraded in fading channels where the amplitude
variation is severe.This receiver provides significant performance improvement in the
BER over the conventional and adaptive MMSE receiver.It uses reduced rank technique,
which project the received signal onto a lower dimensional subspace, thereby resulting
in a covariance matrix with smaller eigenvalue spread.
The successive interference cancellation uses matched filter outputs for esti-
mating the cancellation coefficients and the order of cancellation. After arranging the
signals in order of their strength, the strongest user is first subtracted from the received
signal. [13] This process is repeated for the remaining users, until the weakest user is
remained. In this process, the weakest user had minimum MAI.
The received signals were ranked according to their correlation values, which
were obtained by utilizing the correlation between the received signals and spreading
codes of the users. Each stage of this detector decisions, regenerates and cancels out one
additional directs sequence user from the received signal, so that the remaining users
see less MAI in the next stage. [13] Since symbol estimates associated with a weaker
user, this enables more accurate estimate of the interference. The first stage is preceded
by an operation which ranks the signals in descending order of received powers.
9
ii Parallel Interference Cancellation detector
In an asynchronous system, the outputs from a matched filter bank were fed
into a detector that performed MAI cancellation using a multistage algorithm. At each
stage in the detector, the data estimates of all the other users from the previous stage
were used for reconstructing an estimate of the MAI, and this estimate was then sub-
tracted from the interfered received signal representing the wanted bit. The complexity
of this detector was linear with number of users. The detector has shown better per-
formance for large number of users, where delay introduced by SIC method becomes
intolerable. [13]
Another class of multiuser detectors known as blind detectors, does not require
knowledge of training sequences and CIRs of multiuser interferers. These detectors
do not require the transmission of training sequences or parameter estimates for their
operation. Instead, the parameters are estimated "blindly" according to certain criteria,
hence the term "blind" detection. [13]
This receiver requires only the spreading code of desired user in contrast to
other detectors, which requires spreading codes of other users also.Blind adaptive linear
10
receiver for DS-CDMA systems are promising technique for interference suppression as
they offer an attractive trade-off between performance and complexity and can be used
in situations where a receiver loses track of the desired signal and a training sequence
is not available.A disadvantage of the method is that it suffers from the problem of
signature mismatch and thus cannot work in multipath environments. [13] A related
approach that requires knowledge of all signature waveforms appeared concurrently.
In all the multiuser detectors discussed above, the required parameters except
from the transmitted data were assumed to be known at the receiver.The output of the
channel equalizer is processed by a fixed multiuser detector in order to provide the data
estimates of the desired user.
In multiuser environment, each user in K users system suffers from MAI due
to other (K-1) users. This MAI can also be viewed as if a single users is affected by
Inter symbol interference (ISI) from (K-1) paths in a multipath channel. [13] Therefore,
equalization techniques which are used to remove the effect of ISI can be modified to
remove MAI in multiuser environment. These detectors are known as joint detectors,
which are capable of mitigating the effects of both ISI and MAI.
11
Chapter 2
LITERATURE REVIEW
Much research has been done on the combination of OFDM with multiple access
techniques based on CDMA. Combining CDMA with OFDM yields several schemes:
when the spreading takes place in the frequency domain MC-CDMA is obtained, when
it takes place in the time domain MC-DS-CDMA is obtained. Spreading may also be
applied in both the time and frequency domains.
The concept of using parallel data transmission and Frequency Division Multiplex-
ing (FDM) was first published in the mid 1960s. Some early development can be traced
back to the 1950s [14].In a classical parallel data transmission system, the total car-
rier’s frequency band is divided into N non-overlapping subcarriers and each subcarrier
is modulated by one symbol.
The MC-CDMA transmitter spreads the original signal using a given spreading code
in the frequency domain. More specifically, a fraction of the symbol corresponding to
a chip of the spreading code is transmitted on a distinct subcarrier. For multi-carrier
transmission, it is essential to have frequency non-selective fading over each subcarrier.
Therefore, if the original symbol rate is high enough to become subject to frequency
selective fading, the signal needs first to be serial-to-parallel (S/P) converted before be-
ing spread over the frequency domain. The basic transmitter structure of MC-CDMA is
similar to that of a normal OFDM scheme except that in MC-CDMA the same symbol is
transmitted in parallel through many subcarriers whereas the OFDM scheme transmits
different symbols on every subcarrier [14].
Whether the chips corresponding to one symbol should be placed on adjacent sub-
carriers or not has been discussed in a number of papers [15] [16]. Spreading the symbol
on spaced subcarriers was called chip or carrier interleaving to differentiate it from bit
interleaving applied at the output of the channel encoder. Chip interleaving ensures
that chips of one symbol are transmitted over independent identically distributed sub-
carriers, which increases diversity. On the other hand, it destroys spreading sequence
orthogonality if any leading to a high level of Multiple Access Interference (MAI) in a
multi-code transmission configuration.
In MC-DS-CDMA, the transmitted data stream is S/P converted and then each of
the parallel streams is spread using a given spreading code in the time domain so that
the resulting spectrum of each subcarrier can satisfy the orthogonality condition with a
minimum frequency separation [22]. In [23] MC-DS-CDMA was proposed where each
of the S/P converted streams is repeated on multiple subcarriers after the spreading is
applied to each stream. The effects of two major problems facing MC-DS-CDMA,
frequency offset and phase jitter, were studied by Steendam in [24].
13
be traced back to 1999 when Matsutani [?] proposed MC-DS-CDMA with frequency
domain spreading, in which every transmitted symbol is spread in the frequency domain
over a number of subcarriers before the actual time-domain spreading is applied.
The work reported in [27] gives a fair comparison of frequency domain and time
domain spreading. The paper presented simulation results of the effect of frequency
and time domain spreading factors on the performance of the system in both heavily
and lightly loaded scenarios. The study was conducted using both QPSK and 16QAM
signal sets with the aid of Turbo encoding and receive antenna diversity.
A major feature of MC-CDMA is its ability to exploit the inherent frequency di-
versity of OFDM in frequency selective channels while still dealing with the frequency
selectivity manifested as flat fading per subcarrier. Nevertheless, designing MCCDMA
to achieve the most from frequency diversity yields a new problem: increased inter-code
14
interference across the different codes in use. Therefore, substantial research related to
MC CDMA with focus on minimising or eliminating the intercode interference or MAI
in order to improve the MC-CDMA’s system performance has been carried out. Re-
search in this space can be categorised into Equalisation, Multi User Detection (MUD),
and Interference Cancellation (parallel or successive).
Optimal MUD involves performing a ML search in which the distances between the
received word and each of the possible codeword combinations are calculated to decide
on the one closest to the received signal. ML-MUD is a prohibitively complex process
even for a small number of sequences, and its complexity increases exponentially with
either the increase of number of sequences or the sequence length. Sub-optimal methods
have been proposed in the literature to reduce the number of required searches in the
code space. In [29] pre-whitening tree pruning (PWTP) was proposed as a near-ML
MUD method for MC-CDMA based on the tree search algorithm. Although this method
reduces the complexity of ML detection, it is still complex in itself and quite costly for
hardware implementation as it requires memorising all surviving paths in the search
tree and requires sorting all node metrics at every level of the tree.
Rugini [30], on the other hand, reduced the number of searches for the best esti-
mate in a ML-MUD by applying Local ML (LML). In LML the search is limited to
the neighbourhood of the initial codeword estimate i.e. to all codewords within a cer-
tain Hamming distance from that codeword. This approach significantly reduces the
complexity of the ML decoder. Simulation results however did not mention how close
LML’s performance is compared to the optimal ML MUD.
In 1972, David Chase proposed his new class of algorithms to decode linear block
codes using channel state information [31]. His algorithm gained exceptional interest
from researchers interested in achieving sub-optimal performance not only for linear
block codes, but also for other applications which share the Maximum Likelihood (ML)
algorithm performance as their optimal performance benchmark (the Chase algorithm
is reviewed in more detail in the next section).
It is apparent from the literature that multicarrier modulation is the strongest candi-
15
date to support the requirements of the 4G air interfaces. In terms of the multiple access,
joint OFDM-CDMA schemes inherit the strengths of both OFDM and CDMA. There-
fore, many publications have advocated either MC-DS-CDMA (time domain spreading)
or MC-CDMA (frequency domain spreading) as being the best approach for the multi-
ple access of a multicarrier system. Very few, however, supported their argument with
a fair comparison of both schemes.
16
Chapter 3
Whether the chips corresponding to one symbol should be placed on adjacent sub-
carriers or not is an issue of investigation. Distributing a symbol over the whole avail-
able bandwidth involves chip interleaving and is distinct from bit interleaving applied at
the output of the channel encoder. Chip interleaving ensures that chips of one spreading
code are transmitted over independent identically-distributed subcarriers, increasing the
possibility of having uncorrelated fading over the related subcarriers thereby enabling
the system to exploit more frequency domain diversity. On the other hand, reducing
the correlation in fading among the spreading code’s chips destroys the orthogonality
among the different spreading codes in use which leads to a high level of MAI in a
multi-code transmission scenario. However, when no chip interleaving is applied and
if the total bandwidth occupied by one spreading code is narrower than the coherence
bandwidth of the channel, the channel frequency response becomes flat over that band-
width and there should be no MAI at reception if the spreading codes themselves are
orthogonal. However, in this case no diversity is gained by the receiver in the frequency
domain. Thus it is a trade-off between diversity and MAI. Chip interleaving maximises
the frequency domain diversity exploitation at the expense of increasing MAI while the
opposite maintains MAI at a low level at the expense of frequency diversity.
the intercell interference over a number of subcarriers for each symbol. Whether such
an advantage overcomes the intercell interference caused by MAI is another point of
investigation in order to determine the performance of MC-CDMA.
MAI can be ameliorated with the aid of multi-user detection (MUD). However, good
performing multi-user detectors are always of a considerably high complexity compared
to the basic Single-User detector (SUD). Therefore, improving the performance of MC-
CDMA in terms of detection is frequently at the cost of increasing complexity in the
system. The amount of improvement in the performance of MC-CDMA when moving
from the simple SUD to the more complex MUD is part of this chapter’s investigation.
18
Figure 3.2: MC-CDMA Transmitter Receiver
The MC CDMA is the concatenation of OFDM and CDMA and hence it in re-
sult provides an output of the spreaded signals in N carriers. Each user’s message bits
are first digitally modulated using common modulation techniques like BPSK, QPSK,
16QAM, 64QAM etc. As the number of bits in the constellation diagram of modu-
lation technique increases, the PAPR decreases. The modulated signal is then spread
using spreading sequences as done in the DS CDMA. Walsh- Hadamard codes, gold se-
quence, m-sequences, pn sequences etc are generally used as the spreading sequences.
Good spreading codes need to have an impulsive auto-correlation function, zero cross-
correlation value and a large number of codes should be available to cope with the mul-
tiuser scenario. For simulation purposes, Walsh-Hadamard codes are used as spreading
sequences and the number of carriers are made equal to the length of the spreading se-
quence. Then the signal is serial to parallel converted and IFFT is taken for mapping the
signals to the carriers. A guard interval with a cyclic extension (Cyclic Prefix) is then
added to the signal to immunize it from inter channel interference (ICI). All the users
signals are then summed together and is then carrier modulated. The resultant analog
signal is then transmitted through the antenna.
The time domain representation of the transmitted signal of user k can be given by
19
the equation,
Nc /SF SF
X X
sk (t) = bki (t) cfk (t) e(2πfij t) (3.1)
i=0 j=0
where Nc /SF is the number of symbols transmitted in parallel per subcarrier, bki (t) is
the ith transmitted data symbol of user k at time t , and cfk (t) is the frequency-domain
spreading code of user k at time t
The signal passes through fading channels and are vulnerable to noise and other
interference. The carrier modulation is then removed and the data is converted back to
digital. The cyclic prefix of the obtained signal is then removed and channel equalisa-
tion is applied to the signal. Then FFT is taken and the data is converted to serial. Then
the spreading code itself is used to despread the signal. The applied digital modulation
is then removed by adequate demodulation technique. Then the detection procedure
starts with a single-user detector which is basically a matched filter bank. Then any of
the multiuser detection scheme is used to reconstruct the transmitted bits eliminating
the effects of MAI. Figure 5.8 shows the block diagram of MC-CDMA system with
multiuser detection.
At the receiver, the received signal corresponding to the lth transmitted symbol can
be represented as,
Rl = Xl Hl + Nl = 1, 2, ..., Nc /SF (3.2)
where Xl is an SFf long vector representing the symbols spread over the lth group of
subcarriers in the frequency domain, and is given by
where Cnf (k) is the k th chip of the frequency domain spreading code assigned to the
nth user. bnl is the data symbol corresponding to the nth user and transmitted over the
lth group of subcarriers. The diagonal channel matrix for the signal received subcarrier
group l is Hl and the additive white Gaussian noise vector (AWGN) is given by Nl .
20
If more than one receive antenna are in use, the resultants from all antennas are
maximal ratio combined. The signal is then despread in the frequency domain where
the orthogonality between codes is violated due to the frequency selective fading across
the code chips. The dispreading, detection and separation methods of the user’s signal
fall into one of the following two categories:
Single-User Detection (SUD): This is done the same way as in the time domain by
multiplying the signal with the user’s designated frequency-domain code incorporating
a weighting factor and then summing the resultant signals together. The weighting
factor can take different values depending on the combining method in use; this may
be ORC, MRC, EGC etc. Using a single-user detection technique has the advantage of
simplifying the detection algorithm, and utilising the spreading and frequency diversity
gain, but the MAI issue is not addressed.
Multi-User Detection (MUD): The MUD method used in this chapter is the general
MMSE-based MUD. In this technique the despreading is first done by multiplying the
signal with the frequency-domain code matrix. The next step is to detect the wanted
user signal using MMSE-based MUD which balances the MAI and noise amplification
processes. To obtain the MMSE weight matrix Z , the covariance matrix R is calculated.
After despreading in the frequency domain, the resultant can be sent to the demap-
per; soft outputs are deinterleaved and then forwarded to the channel decoder to estimate
the actual transmitted data bits.
21
Chapter 4
An optimal MUD will consider all the possible variations of symbols that were
spread by each of the spreading codes, regenerate the spread signals, add them together
and compare every combination with the received signal to find the one of minimum
distance from the received signal.Again, the ML detector is extremely complex and
even for short spreading codes its implementation is unrealistic. On the other hand, this
process can be significantly simplified by creating error patterns using only the least
reliable symbols at the output of the detector after an initial estimation using a matched
filter as a SUD. Each of these patterns is then used to reconstruct the transmitted code-
word corrupted by the channel and the results are compared with the received signal.
A decision (soft or hard) is made based on the codeword that is closest to the received
signal. In an uncoded system, this output of the detector can be taken as a final hard
decision to the demapper. If the signal is encoded, a soft output from the detector based
on the reliability metric of the initial SUD detector can be fed to a soft channel decoder.
The main idea of the Chase Algorithm was to simplify the Maximum Likelihood (ML)
search among all possible codewords to a search amongst the most probable alternatives
based on some metric.The Chase Algorithm has two major advantages: Simplicity and
Flexibility.
Simplicity: The Chase algorithm replaces the search along the whole lookup ta-
ble (or code), where every codeword is considered, with a search among a new much
smaller lookup table where only the least reliable positions of a code are replaced with
their alternatives. This dramatically reduces the complexity of the decoder while still
offering a near-ML performance. For example, for a code of base m and length n , the
length of the lookup table is reduced from mn codes for the ML decoder to mp for the
Chase decoder, where p n is the number of least-reliable symbols to be varied.
A detecting scheme for MC-CDMA using chase algorithm with dynamically fixing
the chase factor is proposed. The ml search of all possible keywords is replaced by
the search of most probable alternatives using the chase algorithm [32]. The chase
algorithm replaces the very large all-code look up table of the optimum ml detector to a
flexible number of code words. A near ml performance with a significant reduction in
the complexity is offered by the use of chase algorithm. [33] The length of the lookup
table is reduced from mn of the optimum detector to mp in the chase decoder. The
parameter p is the chase factor and controls the trade off between performance and
complexity. [34] If p = n then the chase decoder resembles the optimum decoder. [35]
The p can be varied from 1 to n thereby increasing complexity and performance. The
steps in the flexible chase algorithm is given by,
Algorithm
Step 1: Initialise the decoding of the test sequence using estimates. Compare it with
pre-loaded test signal information.
Step 2: The chase factor is initialised as the nearest whole number which is 10
percentage of the number of decoded bits.
23
Figure 4.1: Dynamic Chase Algorithm for MUD
Step 3 : Find the number of max errors occurred during initial estimate of the test
signal. Then fix the chase factor as the max number of error occurred if it is greater than
the initialised value of chase factor.
Step 4: The most unreliable p bits are selected according to the reliability matrix
and the corresponding bit positions are recorded.
Step 5: All combinations (2p ) of the p bits are then taken at their corresponding bit
positions.
Step 6: The 2p lookup table is constructed with all the new 2p code words. Then the
euclidean distance of the received vector from each of the vectors in the lookup table is
calculated.
Step 7: The code word which provided the minimum euclidean distance is then
selected as the detected code.
The algorithm is explained with the help of the flowchart given in Fig. 4.1. The
input to the decoder is the equalized signal. The single-user detector is first used in
the chase detection scheme. A test signal can be send first to determine the max num-
24
ber of bit positions where the errors can happen. This information can then be used to
dynamically fix the chase factor. Once the chase factor p is fixed, then a p number of
least reliable bits are selected from the detected matrix. And all code words with the 2p
combinations of the p least reliable bits are created. Then the received signal is recon-
structed using each of the new code words. Then the normal ML detection scheme is
applied on these 2p code words. Then the euclidean distance of each reconstructed sig-
nal from the received signal is calculated and stored. It is then arranged in the ascending
order to find the code word with the least euclidean distance.
The proposed detection scheme thus simplifies the optimum detector. The received
MC-CDMA signal can be written in the matrix form as,
R = SH + N (4.1)
where S is the matrix with elements resulting from the sum of the k th chip of spreading
sequences of all the users. The element of S is given by,
N
X
S(k) = Cn (k)bn (4.2)
n=1
where b is the transmitted symbol and C is the spreading code.H is the diagonal channel
matrix and N is the noise matrix induced by the channel.
The matched filter is first used as the single-user detector, and a first level decoding
takes place there. A test signal decoding process is done prior to the actual signal
decoding. The maximum number of erroneous bit is then estimated from the decoding
results of the test sequence. This is then fixed as the chase factor ’p’ for the chase
decoding.The proposed flexible chase algorithm is shown in the Figure 1.2.
n
ˆ 6= t(i)
X
p= t(i) (4.3)
i=1
After fixing the chase factor, the SUD detected signal is arranged in the ascending
order such that the least reliable bit positions comes first. Then the first ’p’ number
25
of bits are selected and their respective positions are recorded. Then all the 2p combi-
nations of the selected p symbols are calculated to create the new 2p code words.For
example, assume a 4 user case with p = 2. If the SUD detected code is [1010] and the
least reliable bit positions are the first and last, then the new look up table would be,
1 0 1 0
1 0 1 1
[L] =
0 0 1 0
0 0 1 1
For each of the above combinations, the received signal is reconstructed to get the new
lookup table with 2p elements (P ). Then this lookup table is used to calculate the
euclidean distances to the received vector.
n
X
Di = | Pi − Ri |2 (4.4)
i=1
Then the distance matrix D is sorted in ascending order to get the code word with the
minimum distance. Thus the chase decoding effectively decodes the transmitted signal
with ml detection but with reduced complexity.
26
Chapter 5
The simulations are performed on a Windows 8.1 loaded system with 3GB RAM
and Intel i3 processor using MATLAB 7.12.0 (R2011a). For the simulations, eight users
are considered transmitting 1000 bits. Walsh-Hadamard code of length 16 is used as the
spreading sequence. Since spreading factor is made equal to the number of carriers, the
number of carriers Nc is 16. For mapping the symbols to the carrier, a 128 point IFFT is
used. The channel is modeled using a 4-tap Rayleigh channel and then AWGN is added
to the signal.
The effect of MAI in MC-CDMA increases with the increase in the number
of users. For a k-user scenario, the k-1 users provide interference to the kth user. As
the number of users (k) increases, the effect of MAI is prominent. Compared to the
previous simulations with 2 users, the improvement of mud over sud can be clearly
28
Figure 5.3: MC-CDMA Sub-Optimal MUD with Rayleigh Channel
understood from the graph. The MC-CDMA with six users is simulated and then first
a single user detection is performed and the output of those is then used for the mmse
and blind detectors with the help of the R matrix. The BER curve for MMSE and Blind
Detectors along with an SUD in 6 User MC-CDMA is shown in Fig. 5.4.
The Chase algorithm replaces the search along the whole lookup table (or
code), where every codeword is considered, with a search among a new much smaller
lookup table where only the least reliable positions of a code are replaced with their al-
ternatives. This dramatically reduces the complexity of the decoder while still offering
a near-ML performance. For example, for a code of base m and length n , the length
of the lookup table is reduced from mn codes for the ML decoder to mp for the Chase
decoder, where p n is the number of least-reliable symbols to be varied.
In the simulation the p is taken as 2 and the chase algorithm is used to reduce
the sample space of the ml detection based on euclidean distance. The simulation of
MC-CDMA is done with eight users. Each user transmits 1000 bits. PN sequence is
used with goldcodes for the eight users.
29
Figure 5.4: MMSE and Blind Detectors in 6 User MC-CDMA
30
Figure 5.6: Comparison of Proposed Detector with Sub-optimal Detectors with
Rayleigh channel
The performance of the dynamic chase based detector is compared with the single-
user detector and multiuser detector using the BER curve. A matched filter detector
is used as the single-user detector. The multiuser detector is implemented using the
matched filter bank followed by a decorrelating detector. Fig.5.6 compares the BER of
the dynamic chase based detectors with the MUD and SUD.
At a reference of 6 dB SNR, the proposed detector has a BER of 10−2 and the
MMSE detector has a BER of 10−1 . The ber curve comparison of proposed detector
without Rayleigh channel is compared with matched filter detector, mmse detector and
decorrelating detector is shown in Figure 5.7. Here also around 10 times the perfor-
31
Figure 5.7: Comparison of Proposed Detector with Sub-optimal Detectors without
Rayleigh channel
mance of the conventional detector is achieved under the simulation environment with-
out Rayleigh channel. With Rayleigh channel the BER of proposed detector and MMSE
detector is around 10−3 and 10−1.52 at a reference SNR of 8 dB. Figure 5.8 shows the
comparison of BER curve of proposed detector and other conventional detectors with
Rayleigh channel.
32
Figure 5.8: MC-CDMA Multiuser Detection, Single-user Detection and the Proposed
Detector ber curve with Rayleigh Channel
33
Chapter 6
CONCLUSION
The MAI remains an issue for MC-CDMA. Although the use of optimal ML-MUD
to reduce the effects of MAI is theoretically possible, its implementation is prohibitively
complex and costly. The new algorithm is a near-ML detector based on the Dynamic
Chase Algorithm. The proposed algorithm dramatically reduces the MUD processing
complexity compared to the optimum multiuser detector for MC CDMA. The chase also
provides simplicity and flexibility to the trade off between complexity and performance.
The dynamic allocation of the chase factor also provides a significant leap over
the normal chase detection. It will determine the chase factor by sensing the inherent
error occurring in the channel. This fixing with a static upper and lower limits allows a
smooth trade off of complexity and performance.
The complexity of the optimal multiuser detector is reduced by the use of the
proposed flexible chase algorithm. The original ml detector need to calculate the eu-
clidean distances to all 2k code words in the lookup table, whereas while using chase
the lookup table gets reduced to 2p where p is dynamically fixed. The performance is
also better than the sub-optimal detectors. The simulated results shows an appreciable
decrease in the bit error rate while using the proposed detector.
Bibliography
[1] L. Hanzo, L.-L. Yang, E. Kuan, and K. Yen, Single-and multi-carrier DS-CDMA:
multi-user detection, space-time spreading, synchronisation, standards and net-
working. John Wiley & Sons, 2003.
[3] R. Prasad and T. Ojanpera, “An overview of cdma evolution toward wideband
cdma,” Communications Surveys, IEEE, vol. 1, no. 1, pp. 2–29, 1998.
[5] M. Honig, U. Madhow, and S. Verdu, “Blind adaptive multiuser detection,” Infor-
mation Theory, IEEE Transactions on, vol. 41, no. 4, pp. 944–960, 1995.
35
[10] T. Ojanperä, “Overview of multiuser detection/interference cancellation for ds-
cdma,” in Personal Wireless Communications, 1997 IEEE International Confer-
ence on. IEEE, 1997, pp. 115–119.
[11] R. Lupas and S. VerdŰ, “Linear multiuser detectors for synchronous code-division
multiple-access channels,” Information Theory, IEEE Transactions on, vol. 35,
no. 1, pp. 123–136, 1989.
[13] D.-H. Alexandra, J. Holtzman, and Z. Zvonar, “Multiuser detection for cdma sys-
tems,” IEEE Personal Communications (April 1995), pp. 46–58, 1995.
[14] R. Prasad, OFDM for wireless communications systems. Artech House, 2004.
[19] S.-H. Tsai, Y.-P. Lin, and C. J. Kuo, “Mai-free mc-cdma systems based on
hadamard&# 8211; walsh codes,” Signal Processing, IEEE Transactions on,
vol. 54, no. 8, pp. 3166–3179, 2006.
[20] J.-M. Auffray and J.-F. Hélard, “Performance of multicarrier cdma technique com-
bined with space-time block coding over rayleigh channel,” in Spread Spectrum
36
Techniques and Applications, 2002 IEEE Seventh International Symposium on,
vol. 2. IEEE, 2002, pp. 348–352.
[22] L. Hanzo, L.-L. Yang, E. Kuan, and K. Yen, Single-and multi-carrier DS-CDMA:
multi-user detection, space-time spreading, synchronisation, standards and net-
working. John Wiley & Sons, 2003.
37
channels,” in Radio Science Conference, 2004. NRSC 2004. Proceedings of the
Twenty-First National. IEEE, 2004, pp. C34–1.
[29] S.-F. Tsai, Y.-L. Li, and Y. Lee, “Relative threshold tree pruning multi-user de-
tection with user ordering and user partitioning for mc-cdma,” in IEEE Wireless
Communications and Networking Conference, 2004, pp. 48–53.
[31] D. Chase, “Class of algorithms for decoding block codes with channel measure-
ment information,” Information Theory, IEEE Transactions on, vol. 18, no. 1, pp.
170–182, 1972.
[32] Y. Wu, “Fast chase decoding algorithms and architectures for reed–solomon
codes,” Information Theory, IEEE Transactions on, vol. 58, no. 1, pp. 109–129,
2012.
[34] T. Yang and H. Fu, “Approximate ml detection for mimo-ofdm systems using
multistage chase detection algorithm,” in Wireless Communications Networking
and Mobile Computing (WiCOM), 2010 6th International Conference on. IEEE,
2010, pp. 1–4.
38